Temporal Raises $550M Series E at $12.55B, 2.5x in 7 Months

Temporal Technologies just raised $550 million at a $12.55 billion valuation. The round was co-led by Lightspeed Venture Partners, Wellington Management, Growth Equity at Goldman Sachs Alternatives, and Tiger Global, with T. Rowe Price and SV Angel joining and returning backers a16z, Sequoia, Index Ventures, GIC, Sapphire Ventures and Amplify. The round size isn't the interesting number. The multiple is: Temporal was valued at $5 billion in a $300 million Series D back in February. Seven months later it's worth $12.55 billion, a 2.5x markup, because AI agents keep crashing in ways regular software mostly didn't.

Why do AI agents need durable execution?

An AI agent isn't a single function call. It's a loop: call a model, call a tool, wait for a response, call another tool, maybe wait hours for a human to approve something, then keep going. That loop can run for minutes or days, and at any point in the chain a server can restart, a network call can time out, or a process can just die. Traditional apps handle this by writing progress to a database and hand-rolling retry logic at every step. That works until the app has fifty different places where state needs to be saved and fifty different ways state can drift out of sync with reality.

RelatedWonderful's $550M Series C Doubles Valuation to $5B

Temporal calls its approach durable execution. Instead of trusting a single running process to hold the state of a long task in memory, it writes every step, every tool call, every wait, to an append-only event history. If a worker dies mid-task, a new worker doesn't start over. It replays the event history, deterministically reconstructing the exact state the workflow was in, and resumes from that point. No re-running side effects, no lost context. That's the mechanism most coverage skips: enterprises are paying for this now because agent loops turned a niche infrastructure problem into something every company shipping an agent product hits in week one.

What's driving the growth numbers?

The usage data backs up the valuation jump. Temporal's annualized revenue run rate recently passed $250 million, up more than 200% year over year. Net dollar retention has stayed above 200% since February, meaning existing customers are expanding their usage faster than new logos are even coming in. The platform processed 1.9 trillion billable actions in August alone, up 350% year over year, and open-source installs have crossed 43 million, up 134% since December 2025. Paying customers are past 4,300, up 139% year over year, and headcount has doubled to 570 employees in the past twelve months.

The named customers tell the same story from a different angle. OpenAI's own usage of Temporal grew 60-fold in under a year, which is a striking number for a company that builds its own infrastructure by default. Snap runs 414 million Stories a day through Temporal. JPMorgan Chase runs it in regulated production, not a sandbox. Nvidia is on the list too. This isn't a tool being tried out in a hackathon; it's sitting underneath systems where downtime has a dollar cost attached.

How does Temporal compare to Step Functions, Inngest and Restate?

Durable execution isn't a category Temporal invented alone, though it did originate the idea inside Uber's Cadence project back in 2019. AWS Step Functions, Inngest, Restate and Dapr Workflows all solve some version of the same problem: keep a long-running, multi-step process alive across failures. Where they differ is how much control and portability you get.

Temporal's pitch has always been that workflows are written as ordinary code in a real language, not a JSON state machine, and that you can run the same engine on your own servers or hand it to Temporal Cloud without rewriting anything. That portability is part of why JPMorgan can run it in a regulated environment while OpenAI runs it at hyperscale.

RelatedGimlet Labs Raises $300M at $3B to Bet Against Nvidia Lock-In

1. 2019Temporal founded by Samar Abbas and Maxim Fateev, evolved from Uber's Cadence project
2. 2026-02-17Series D: $300M at $5B valuation already reflected early agent-infrastructure demand
3. 2026-09-14Series E: $550M at $12.55B co-led by Lightspeed, Wellington, Goldman Sachs Alternatives, Tiger Global

What does this mean for the valuation?

The signal for investors is straightforward: net dollar retention above 200% and a 60x usage jump from a company like OpenAI mean customers aren't just adopting Temporal, they're becoming structurally dependent on it as agent workloads scale. That justifies premium pricing on a follow-on round. The risk is just as straightforward: a $12.55 billion valuation against roughly $250 million in annualized revenue works out to near a 50x revenue multiple, aggressive even against the current run of AI infrastructure rounds. That multiple only holds up if growth keeps compounding at close to its current pace for several more quarters.

Our take

CEO Samar Abbas put it well: reliability has never been optional, but AI quickly raised the cost of skipping it. Agent loops fail constantly right now, tool calls time out, models hang, servers restart mid-task, and every company shipping an agent product has to answer what happens when that happens at 3 a.m. Temporal turned that question into a business by being the boring layer nobody thinks about until it saves their weekend. What's notable isn't that infrastructure companies are raising big money during an AI boom, plenty are. It's that retention above 200% and usage up 350% year over year are the kind of numbers you get from a company customers can't easily rip out. That's a stickier position than most AI-adjacent startups can claim, worth more of the premium than a simple revenue multiple suggests, even if 50x leaves little room for a stumble.

Original analysis by GenZTech. Source: Temporal.